Technical SEO is not a separate discipline. It is the set of conditions that let the other work be seen.

Crawl budget, in plain terms

Search engines allocate a limited amount of fetching to each site. On a small site you will never notice. On a site with tens of thousands of URLs, it decides what gets refreshed.

You increase effective crawl capacity by:

  • Removing URLs that should not exist (faceted combinations, tracking parameters, infinite calendars).
  • Returning 404 or 410 for pages that are genuinely gone, rather than redirecting everything to the homepage.
  • Keeping important pages shallow in the link graph.
  • Serving fast, stable responses. Timeouts and 5xx responses waste crawl.

You do not increase it by adding more sitemaps.

Status codes that matter

  • 200. Fine.
  • 301. A permanent move. Use it when a URL has genuinely changed. Update internal links to point at the new URL so you are not relying on the redirect forever.
  • 302. Temporary. Rarely what you want for a permanent restructure.
  • 404. Not found. Correct for deleted pages.
  • 410. Gone, permanently. Slightly faster to drop from the index than a 404.
  • 5xx. Server failure. Crawlers back off. Persistent 5xx is the fastest way to lose visibility.

The most common self-inflicted wound is a redirect chain: A to B to C. Each hop costs a request. Collapse chains to a single hop.

Canonicals

A canonical tag says "this is the version that counts". It is a hint, but a strong one.

  • Every indexable page should have a self-referencing canonical with an absolute URL.
  • Duplicate content should point at the original, not at each other.
  • Never canonicalise every page to the homepage. That tells the engine every page is a duplicate of one page, and the rest can disappear from the index.
  • Canonical and hreflang must agree. A common international bug is a canonical pointing at the wrong locale.

Robots and sitemaps

robots.txt controls fetching. A meta robots tag or an X-Robots-Tag header controls indexing. They are different tools and people mix them up constantly.

  • Use robots.txt to keep crawlers out of admin areas, search results and parameter sprawl.
  • Use noindex when a page can be fetched but should not appear in results.
  • Never combine "noindex" with a robots.txt block. If the crawler cannot fetch the page, it cannot see the noindex instruction.

Sitemaps should list canonical, indexable, 200-status URLs with a lastmod you actually maintain.

Speed and Core Web Vitals

Core Web Vitals are a tiebreaker, not a magic lever. Still, a slow page loses readers, and readers are the point.

  • LCP (largest contentful paint) is usually an image or a hero heading. Serve it early, compress it, do not lazy-load it.
  • INP (interaction to next paint) is about main-thread work. Ship less JavaScript.
  • CLS (cumulative layout shift) is about reserving space for images, embeds and fonts.

Measure with field data (real users), not just lab tools. A page can pass Lighthouse and fail in the field.

JavaScript SEO

If the page is built in the browser:

  • Ensure the main content, headings, links and metadata exist in the server response.
  • Avoid rendering canonical tags or structured data client-side only.
  • Test with a rendered fetch, not your browser.

A useful test: disable JavaScript, load the page, and read it. If you cannot tell what the page is about, neither can a crawler on its first pass.

The audit order

When someone asks you to "do a technical audit", work in this order:

  1. Indexability: is anything important blocked, noindexed or canonicalised away?
  2. Crawl paths: are important pages linked and shallow?
  3. Status codes: are there chains, loops, soft 404s?
  4. Duplication: are there multiple URLs for one page?
  5. Rendering: is the content in the HTML?
  6. Speed: what is slow, for whom, on what device?
  7. Structured data: does it validate and match the page?

That order matters. Fixing speed on a page that cannot be indexed is a waste of a week.